In the speech-to-song illusion certain spoken phrases are perceived as sung after repetition. Prior work has established that perception of the illusion is not limited to specialist listeners, such as individuals with musical training, but is perceived widely across the general population. However, little is known about whether there are stable individual differences in perception of the illusion, and if so, what factors underlie this variability. Here we assessed the strength of the song illusion across individuals using phrases that tend to be perceived as song when repeated, as well as phrases that continue to be perceived as speech when repeated, measuring the strength of the illusion as the rating difference between these two stimulus categories after repetition. Illusion strength varied widely, with differences in perception of the two categories ranging from 0% to 80% of the rating scale. Although variability in illusion strength was unrelated to degree of musical training, participants who perceived the illusion more strongly were proficient in several musical skills, including beat perception, tonality perception, and selective attention to pitch. This finding supports theoretical models of the speech-to-song illusion in which experience of the illusion is based on detection of musical characteristics latent in spoken phrases.